Programmable Structured Synthetic Data Generation for Privacy-Preserving Data Sharing​

This project proposes the development of a programmable synthetic data generation framework to enable privacy-preserving data sharing. 

By leveraging hybrid NLP models and Generative AI, the algorithm aims to generate structured synthetic datasets that statistically resemble real-world PII data while ensuring data privacy. The project primarily addresses the issue of fairness in AI model training. 

Project Deliverables/Outcomes/Impact:

• A hybrid NLP module for accurate identification of PII in structured datasets.
• A GAN-based synthetic data generator integrated with DP for enhanced privacy.
• A benchmarking tool to assess utility, fairness, and privacy risk of synthetic datasets.